A Lagrangian 1-year climatology of (deep) cross-tropopause exchange in the extratropical Northern Hemisphere

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 107, NO. D2, 4021, /2001JD000812, 2002 A Lagrangian 1-year climatology of (deep) cross-tropopause exchange in the extratropical Northern Hemisphere Heini Wernli and Michel Bourqui 1 Institute for Atmospheric and Climate Science, Eidgenossische Technische Hochschule, Zürich, Switzerland Received 7 May 2001; revised 6 September 2001; accepted 10 September 2001; published 31 January [1] A Lagrangian methodology is applied to operational European Centre for Medium-Range Weather Forecasts analyses to study upward cross-tropopause exchange (troposphere to stratosphere exchange (TSE)) and downward cross-tropopause exchange (stratosphere to troposphere exchange (STE)) in the extratropical Northern Hemisphere for the period from May 1995 to April A residence time criterion serves to distinguish between short- (<1 2 days) and long-lasting exchange events. The adopted approach enables identification of a range of novel aspects of extratropical cross-tropopause transport which are of primary importance when assessing its chemical impact. For the considered year the annual cycle of the hemispherically integrated net cross-tropopause mass flux compares well with estimates from previous studies. The part of STE and TSE which occurs with equal amplitude in both directions (referred to as symmetric two-way exchange ) has only a weak annual cycle and, for short residence times, a larger amplitude than the net exchange. The meridional distribution of the net flux reveals an upward branch in the subtropics, pronounced downward exchange in the mids and weak upward fluxes in the Arctic region. Detailed geographical distributions show significant zonal asymmetries with maximum exchange in the Atlantic and Pacific storm track regions. It is further shown that STE (TSE) events occur typically below (slightly above) the climatological tropopause position. Deep exchange (that is, rapid vertical transport between the stratosphere and the lower troposphere) is strongest during winter and confined to the mid regions of baroclinic wave activity. The localized source regions for deep TSE indicate that pollutants emitted in eastern North America and Asia have an enhanced potential for being rapidly transported into the lowermost stratosphere. INDEX TERMS: 3362 Meteorology and Atmospheric Dynamics: Stratosphere/ troposphere interactions; 3364 Meteorology and Atmospheric Dynamics: Synoptic-scale meteorology; 0368 Atmospheric Composition and Structure: Troposphere constituent transport and chemistry; KEYWORDS: stratosphere-troposphere exchange, vertical transport, tropopause dynamics, trajectory analysis 1. Introduction [2] Stratosphere-troposphere exchange is one of the subjects of the atmospheric sciences which involves many subdisciplines and a wide range of spatial and temporal scales. The quantitative assessment of the exchange fluxes of air, water vapor, ozone, and other chemical constituents is of major importance for atmospheric chemistry (both in the stratosphere and the troposphere) and climate. For example, Butchart and Scaife [2001] indicate the possibility of an increased mass exchange between the stratosphere and troposphere in a changing climate, which could have a significant impact on the distribution of atmospheric constituents. From a more qualitative standpoint it is a challenging task to identify the dynamical and physical processes that are responsible for the associated large-scale transport and for the formation and dissipation of the mesoscale structures near the tropopause where the exchange events actually occur. Because of the large variety of involved aspects many different approaches and methods have been applied in past studies of cross-tropopause exchange. They can be broadly classified into (1) climatological global-scale studies and (2) detailed synoptic-scale case studies. In the next 1 Now at Department of Meteorology, University of Reading, Reading, UK. Copyright 2002 by the American Geophysical Union /02/2001JD paragraphs a brief summary is provided of the approaches, questions, and results of the two categories. [3] In a review article on stratosphere-troposphere exchange, Holton et al. [1995] discussed the global-scale residual mean meridional circulation and described the basic dynamical processes that are likely to drive this circulation if viewed within a twodimensional (zonally averaged) framework. Juckes [1997, 2000] focused on the tropopause region and derived estimates of the mean meridional mass flux based upon quasi-geostrophic theory. Using UK Meterological Office (UKMO) analysis data, Appenzeller et al. [1996b] combined the annual cycles of the residual mean circulation and of the mass of the lowermost stratosphere to calculate the hemispherically integrated net downward cross-tropopause mass flux in the extratropics. The results from this study serve as a milestone for quantitatively validating the outcome of other techniques. Among these, the Eulerian approach introduced by Wei [1987] has been the most frequently applied. It can be formulated with different vertical coordinates (height, pressure, potential temperature (), potential vorticity (PV); see Wirth and Egger [1999] for a careful discussion) and, in essence, estimates the motion of air relative to the tropopause. Hoerling et al. [1993], Grewe and Dameris [1996], and Siegmund et al. [1996] used the approach of Wei [1987] (with pressure or potential temperature as vertical coordinate) to investigate the net cross-tropopause mass flux for selected months or 10 winter seasons [Grewe and Dameris, 1996]. However, Wirth and Egger [1999] and Gettelman and Sobel [2000] pointed out that the Wei approach (except for its formula- ACL 13-1

2 ACL 13-2 WERNLI AND BOURQUI: EXTRATROPICAL CROSS-TROPOPAUSE EXCHANGE tion with PV as vertical coordinate) suffers from cancellation problems (near compensation of large terms) and is highly sensitive to inaccuracies in the underlying data. [4] In the second category the focus has been more on the study of archetypal synoptic and mesoscale weather patterns (and tropopause structures) that accompany cross-tropopause exchange events and on the underlying physical processes. These studies are usually based upon relatively high resolution data from in situ and remote observations, operational analyses or mesoscale model simulations. Here a selection of such investigations is discussed in order to illustrate the variety of analyzed meteorological structures and of the applied methodologies. Lamarque and Hess [1994] used a mesoscale model simulation of a deep tropopause fold associated with a mid cyclone to estimate the cross-tropopause mass flux within this system on a timescale of 4 days. A PV budget analysis indicated that the total mass exchange was dominated by diabatic processes but that the relative importance of turbulent diffusion could be significantly case dependent. The evolution of narrow and comparatively shallow stratospheric filaments was considered by Appenzeller et al. [1996a]. On the basis of water vapor satellite imagery and contour advection calculations they provided evidence for mixing of stratospheric and tropospheric air masses during the breakup of stratospheric streamers. Such a breakup process was studied in detail by Bourqui [2001], who quantified the associated exchange mass fluxes and pointed out the complex variety of the involved physical processes, even for such a seemingly simple system. The radiative effects on dry stratospheric filaments were investigated in isolation in the idealized study by Forster and Wirth [2000]. O Connor et al. [1999] analyzed a filamentary structure of tropospheric air and found significant upward cross-tropopause mass flux associated with the transport of subtropical air toward northern Europe. Finally, the decay of cut-off cyclones has been considered, for instance, by Wirth and Egger [1999] and Gouget et al. [2000]. Gouget et al. showed that in addition to direct convective erosion, filamentation of the outer layers can be an important contribution to the system s decay. Wirth and Egger [1999] focused on methodological issues and compared mass exchange estimates on the basis of different techniques. Trustworthy results have been obtained only with the Lagrangian approach and the Wei method with PV as vertical coordinate. These methods were also used by Kowol-Santen et al. [2000], who found that the two methods yield similar values for the net cross-tropopause mass flux. However, the separate upward and downward fluxes were significantly larger when using the Eulerian approach. [5] In this paper we apply a fully Lagrangian (i.e., trajectory based) technique to study extratropical cross-tropopause exchange for a time period of 1 year on a hemispheric scale. Using comparatively high resolution T213L31 analysis data from the European Centre for Medium-Range Weather Forecasts (ECMWF), this study aims at reducing the gap between the two main approaches mentioned above. Already pioneering studies on the subject used air parcel trajectories (and, in particular, the material change of PV along them) to identify and analyze cross-tropopause exchange [Reed, 1955; Staley, 1962; Danielsen, 1968]. Trajectories are also of central importance for the interpretation of measurements which provide observational evidence for cross-tropopause exchange. Recent examples are the analysis of surface observations of ozone and beryllium isotopes in the Alpine area [Stohl et al., 2000], Lidar measurements in central and southern Europe [Stohl and Trickl, 1999; Kowol-Santen and Ancellet, 2000], and ozone soundings from the Canary Islands [Kentarchos et al., 2000]. To obtain quantitative estimates of exchange mass fluxes, several trajectory-based techniques have been developed in the last years. Sigmond et al. [2000] and Meloen et al. [2001] promoted a quasi-lagrangian technique where PV changes along trajectories served to estimate the material time derivative in the Wei formula with PV as vertical coordinate. This technique can handle folded tropopause structures (unlike the purely classical Wei approach) but still has technical problems in regions with an almost vertical tropopause. The fully Lagrangian methodologies of Wernli and Davies [1997] and Stohl [2001] are independent of the geometrical shape of the tropopause. Stohl [2001] considers the transport across a finite-width tropopause layer, whereas the key characteristic of the study of Wernli and Davies [1997] is the investigation of the exchange parcel s residence time in the stratosphere and troposphere before and after the crossing of the tropopause and of the extent of vertical transport in the troposphere. It is this technique that is extended and applied in the present study (see section 2). 2. Methodology [6] The Lagrangian quantification of cross-tropopause exchange is based upon trajectory calculations with the Lagrangian Analysis Tool (LAGRANTO) [Wernli and Davies, 1997] and operational ECMWF T213L31 analysis data for the 1-year time period May 1995 to April During this time period, at the end of January 1996, the three-dimensional variational analysis technique was made operational at the ECMWF which, according to Morgenstern and Carver [2001], led to an improvement of the data quality in the tropopause region. The analysis data have been interpolated on a regular longitude- grid with a horizontal resolution of 1 and resolve the relevant (sub) synoptic-scale dynamical structures including stratospheric filaments and tropopause folds. Secondary variables like and PV have been calculated directly on the hybrid model levels. The resolved winds of the ECMWF model that are used for the trajectory computations do not explicitly resolve the intense vertical motions associated with deep convection, and this has an impact upon the accuracy of the trajectories in the troposphere. This issue will be further discussed in section 4.1. [7] To identify exchange events either from the stratosphere to the troposphere (STE) or vice versa (TSE) the time traces of PV along a large set of trajectories are analyzed. Using the traditional dynamical tropopause definition, i.e., the 2-pvu isosurface (pvu denotes the potential vorticity unit and corresponds to 10 6 m 2 s 1 Kkg 1 )[Hoskins et al., 1985], the following three-step procedure is applied: 1. At 0000 UTC on every day of the 1-year period, trajectories are calculated over a time period of 24 hours starting from a regular grid covering the whole Northern Hemisphere in between 590 and 50 hpa. The grid has increments of 80 km in the horizontal and 30 hpa in the vertical, and every trajectory represents an air mass of m = g 1 xyp kg, where g is the gravitational acceleration. The resulting number of trajectories is per day and for the whole year. 2. Each 24-hour trajectory is examined whether it crosses the 2- pvu surface once during this time period. These events are considered as preliminary STE or TSE events, and the location where the 2-pvu surface is crossed is taken as the exchange location. 3. The preliminary 1-day exchange trajectories are extended both 4 days forward and 4 days backward in time which yields 9- day trajectories that are symmetric in time around the time of exchange. Then the air parcel s residence times in the stratosphere, t s, and in the troposphere, t t, are determined, and air parcels where both residence times are larger than a threshold value t 8 are selected as significant exchange events. [8] The residence time concept is illustrated schematically in Figure 1. Shown are three trajectories all with a long residence time in the stratosphere (t s ) but with strongly differing tropospheric

3 WERNLI AND BOURQUI: EXTRATROPICAL CROSS-TROPOPAUSE EXCHANGE ACL 13-3 τ s Figure 1. A schematic vertical section showing the stratosphere (shaded area), troposphere (open area), and three prototype trajectories which intersect the tropopause (see text for details). residence times (t t ). Trajectory 1 only remains very briefly in the troposphere before reentering the stratosphere. Therefore the tropopause crossings labeled ST 1 and TS 1 are not considered as significant exchange events, and they are excluded by our selection criterion. In contrast, trajectories 2 and 3 remain for a considerably long time in the troposphere and the crossings labeled ST 2 and ST 3 are considered significant exchange events. Trajectory 3 reaches into the lowermost troposphere, and hence the exchange event labeled ST 3 is considered a deep exchange event. [9] This residence time criterion is introduced mainly to eliminate air parcels which move rapidly to and fro across the tropopause and which have possibly a negligible chemical impact. It also makes an allowance for the anticipated errors in calculating the parcel trajectories and PV in the vicinity of the richly structured extratropical tropopause and avoids the consideration of spuriously identified exchange events. In this study, STE and TSE mass fluxes have been calculated for threshold residence times t 8 = 24, 48, 72, and 96 hours, and the focus will τ t be on the results with the 96-hour value, i.e., on the almost irreversible events. A more thorough discussion of the methodology and a detailed application to a case study is provided by Bourqui [2001]. [10] The application of this methodology leads to comprehensive separate data sets of upward and downward exchange events, with detailed information about the time and location of the exchange and about the 4-day history and future of the exchange air parcels. From these data sets, gridded monthly distributions of TSE and STE are produced with a horizontal resolution of 3. An important distinction will be made between the net (STE-TSE) and two-way exchange flux, which refers to the mass flux that occurs with equal amplitude in both directions within a certain bin (calculated as min(ste,tse)). Although this two-way exchange has no effect on the mass budget of the troposphere and stratosphere, it may have a significant impact upon chemistry because of the differing chemical compositions of upward and downward fluxes. In section 3 the results are presented with increasing geographical detail, starting with the annual cycle of the hemispherically integrated cross-tropopause flux in section 3.1. Section 3.2 portrays the meridional distribution, and the preferred geographical and vertical exchange locations are discussed in sections 3.3 and 3.4. Finally, section 3.5 focuses on the special category of vertically deep exchange events. In all sections, consideration is given to the sensitivity to the choice of the threshold residence time. 3. Cross-Tropopause Mass Fluxes 3.1. Annual Cycle [11] Figure 2 illustrates the annual cycles of the horizontally integrated net and two-way extratropical cross-tropopause mass fluxes for different residence time threshold values (the horizontal integration extends over the region north of 20 N). For the net flux, several notable features are evident from Figure 2a: (1) The general structure of the annual cycle (largest values from January to April and smallest values from August to October) is fairly similar for all values of t 8 ; (2) for t 8 = 24 hours (i.e., if very short lived exchange events are also taken into account) the net mass flux values become negative (i.e., upward) during the summer and autumn period; (3) for longer residence times (48 96 hours) the curves are almost identical and positive throughout the year, with a flat maximum in winter/spring (of kg s 1 ) and about 9-1 net exchange mass flux [10 kg s ] a) STE-TSE month 9-1 two-way exchange mass flux [10 kg s ] b) month min(ste,tse) Figure 2. Annual variation of (a) the net downward mass flux and (b) the two-way exchange flux (in units of 10 9 kg s 1 ) across the Northern Hemisphere s extratropical tropopause (i.e., the region N) for different residence time threshold values. The solid, dashed-dotted, long-dashed, and short-dashed curves are for residence times t 8 = 96, 72, 48, and 24 hours, respectively.

4 ACL 13-4 WERNLI AND BOURQUI: EXTRATROPICAL CROSS-TROPOPAUSE EXCHANGE downward mass flux [10 kg km s ] a) STE upward mass flux [10 kg km s ] b) TSE net exchange mass flux [10 kg km s ] c) STE-TSE two way exchange mass flux [10 kg km s ] d) min(ste,tse) Figure 3. Zonally integrated cross-tropopause mass fluxes for the four seasons from May 1995 through April 1996 based upon exchange trajectories with a threshold residence time of 96 hours: (a) stratosphere to troposphere exchange (STE), (b) troposphere to stratosphere exchange (TSE), (c) net (STE-TSE), and (d) the two-way exchange component (min(ste,tse)). Values are in 10 6 kg km 1 s 1. The solid line is for winter, the long-dashed line is for spring, the short-dashed is for summer, and the dashed-dotted line is for autumn. 3 4 times smaller values from July to October. Note that this net downward mass flux in the extratropical regions must be balanced by an equally strong upward mass flux in the tropics (which is not explicitly considered in this study). The exchange events with t 8 = 96 hours are also contained in the t 8 = 24 hours ensemble, and the difference of the two curves corresponds to the net cross-tropopause mass flux of the exchange events with residence times in between 1 and 4 days. This flux is roughly zero in spring and negative throughout the rest of the year with a peak value of about kg s 1 in October. The sensitivity of the net mass flux to the residence time threshold value is related to an asymmetry between STE and TSE events, as further illustrated in the section 3.2 and explained in detail in Appendix A. [12] The annual cycle with t 8 = hours depicted in Figure 2a, in particular the overall shape and the amplitude of the seasonal variation, compare quite well with the results of Appenzeller et al. [1996b] for the 2 years 1992 and The absolute values differ by some 20 40% (if we restrict the horizontal integration to the domain north of 28 N, then this difference becomes almost zero [Bourqui, 2001]), and considering the completely different methods, the resolution of the underlying data sets, and the possibility of interannual variability, the correspondence is encouraging. (The UKMO data used by Appenzeller et al. [1996b] have a horizontal resolution of and six levels between 50 and 300 hpa. For the ECMWF data we used these values are 1 for the horizontal resolution and 11 levels for the same pressure range.) For the winter season, Gettelman and Sobel, [2000, Figure 8] provide a compilation of the net Northern Hemisphere extratropical mass flux values from different studies at various levels. In comparison with the other existing estimates for the 2-pvu tropopause, our wintertime mean value of kg s 1 is slightly smaller than the values of Gettelman and Sobel [2000] ( kg s 1, mass budget approach), Appenzeller et al. [1996b] ( kg s 1, mass budget approach) and Stohl [2001] ( kg s 1, Lagrangian approach), and they are all significantly smaller than the value of Siegmund et al. [1996] ( kg s 1 ) which is based upon the Wei formula. Note, however, that these values are sensitive to the choice of the tropopause level; our method yields a slightly enhanced wintertime net mass flux value of kg s 1 for the 1.5-pvu surface

5 WERNLI AND BOURQUI: EXTRATROPICAL CROSS-TROPOPAUSE EXCHANGE ACL 13-5 irreversible exchange yield comparable net flux estimates but much smaller values for STE and TSE [Wernli and Davies, 1997; Stohl, 2001]. Figure 4. Comparison of zonally integrated (thin lines, in units of 10 6 kg km 1 s 1 ) and zonally averaged (bold lines, in units of 25 kg km 2 s 1 ) cross-tropopause mass fluxes for the winter months December 1995 to February 1996 based upon exchange trajectories with a threshold residence time of 96 hours. Solid lines are for STE, and dashed lines are for TSE. and a reduction to kg s 1 for the 3-pvu surface, as discussed in more detail by Bourqui [2001]. [13] In contrast to the net cross-tropopause mass flux, the twoway flux reveals almost no annual cycle and a pronounced sensitivity to the residence time threshold value (Figure 2b). If short-lived exchange events (t 8 = 24 hours) are also included, then the amplitude of the annual mean two-way flux is about an order of magnitude larger than the net exchange. For long-lived events (t 8 = 96 hours) the ratio is smaller, and the net flux becomes comparable in magnitude to the upward and downward fluxes. This shows that for short residence time thresholds (or for Eulerian methods [Lamarque and Hess, 1994]) the net flux is the difference of two large numbers, whereas Lagrangian techniques with a more severe threshold criterion for significant almost 3.2. Meridional Distribution STE and TSE. [14] Figure 3 shows the meridional distribution of the zonally integrated cross-tropopause mass fluxes (with a threshold residence time of t 8 = 96 hours) for the four seasons (spring corresponds to March, April and May, etc.). STE (Figure 3a) is most intense in the mids between 35 and 50 N, whereas TSE (Figure 3b) shows relatively large values in the subtropics (20 30 N), a weak minimum near N, and again larger values in the northern mids (45 60 N). Generally, the cross-tropopause mass fluxes are largest in winter and smallest in summer. The most pronounced seasonal variability occurs in the southern part of the mids (30 45 N) where the STE winter values are 2 3 times larger than the summer ones. In comparison, the seasonal differences for TSE are moderate in the entire hemisphere. [15] It is important to point out that the zonally integrated mass fluxes (compare Figures 3a and 3b) are relevant, for instance, for assessing the relative importance of certain bins for global budgets and that the calculation of zonally averaged mass fluxes is more appropriate for investigating regional processes. The two quantities differ by a factor proportional to cos l (where l denotes ) and yield a different impression of the relative importance of low versus high s, in particular for TSE (Figure 4). The zonally integrated TSE mass flux has largest amplitude in the mids, whereas the zonally averaged TSE mass flux peaks in the Arctic region. This shows that (outside the tropics) local TSE mass fluxes are maximum in the high s but that the largest contribution to the total extratropical TSE flux comes from the region near N. For STE both the zonally integrated and the averaged mass flux show a maximum near N. When considering qualitatively the meridional net exchange distribution (STE-TSE; see section 3.4), the distinction between zonally integrated and averaged fluxes becomes less important because the regions with positive and negative values are identical for both quantities Net and two-way exchange. [16] The zonally integrated net exchange (STE-TSE; Figure 3c) is upward (i.e., negative) in the subtropical belt south of N in all four 12 a) STE 12 b) TSE 15-1 downward mass flux [10 kg (2K) ] upward mass flux [10 kg (2K) ] potential temperature potential temperature Figure 5. Potential temperature distribution of (a) STE and (b) TSE events for the four seasons from May 1995 through April 1996 based upon exchange trajectories with a threshold residence time of 96 hours. The seasonally integrated values are in kg (2 K) 1 ; the solid line is for winter, the long-dashed line is for spring, the shortdashed line is for summer, and the dashed-dotted line is for autumn.

6 ACL 13-6 WERNLI AND BOURQUI: EXTRATROPICAL CROSS-TROPOPAUSE EXCHANGE downward mass flux [10 kg km s ] a) STE upward mass flux [10 kg km s ] b) TSE net exchange mass flux [10 kg km s ] c) STE-TSE two way exchange mass flux [10 kg km s ] d) min(ste,tse) Figure 6. The zonally integrated cross-tropopause mass fluxes (for the winter season) for different threshold residence times t 8 : (a) STE, (b) TSE, (c) net downward exchange STE-TSE, and (d) two-way exchange. As in Figure 2 the short-dashed, long-dashed, and solid lines correspond to t 8 = 24, 48, and 96 hours, respectively. seasons. The region with net downward exchange is more variable during the year with a maximum peak near 40 N in winter and a significant decrease in amplitude and a northward shift of the peak to about 50 N in summer. In the polar region, net exchange is weakly upward during the entire year. The zonally integrated twoway exchange distribution (Figure 3d) reveals a fairly simple meridionally symmetric pattern with maximum values in the mids near 50 N. In contrast to the net exchange, the seasonal variability of the two-way exchange is weak with only slightly enhanced values during winter (as previously seen in Figure 2b). [17] Note that the identified meridional distribution of the net cross-tropopause transport differs substantially from the estimates based upon the Eulerian Wei approach [Hoerling et al., 1993; Grewe and Dameris, 1996; Siegmund et al., 1996], where net downward exchange is mainly restricted to the region south of 40 N. The reason for this discrepancy might be associated with the caveats of the Wei approach mentioned in section 1. The net upward exchange in the Arctic region qualitatively agrees with the theoretical considerations of Juckes [2000] Potential temperature distributions. [18] It is also revealing to consider the distributions of the potential temperature values of the individual exchange events (Figure 5) and to compare them with the meridional distributions shown in Figures 3a and 3b. During winter most STE events occur in the range from 285 to 320 K (in agreement with the airborne observations by Zahn [2001]) and almost none above 340 K (Figure 5a). This distribution is significantly shifted toward larger values during summer: Here the maximum values lie between 310 and 340 K, and some events have a value larger than 350 K. Note that the reason for this shift is the annual cycle of the vertical location of the isentropes and not a shift of the exchange activity to more southern s (compare Figure 3a). For TSE the picture is very similar (Figure 5b). The winter distribution peaks near K and shows a distinct minimum at 340 K, whereas during summer, exchange activity is large near K and at 360 K. Again this shift in potential temperature is not reflected as a marked shift in the meridional distribution (see Figure 3b). Qualitatively, these results compare well with isentropic transport studies (in contrast to the present approach, isentropic transport studies are based upon

7 WERNLI AND BOURQUI: EXTRATROPICAL CROSS-TROPOPAUSE EXCHANGE ACL 13-7 STE winter STE spring a) b) STE summer STE autumn c) d) Figure 7. Geographical distribution of the STE mass flux for the four seasons from May 1995 through April 1996 based upon exchange trajectories with a threshold residence time of 96 hours: (a) winter (December, January, February), (b) spring (March, April, May), (c) summer (June, July, August) and (d) autumn (September, October, November). Values are in kg km 2 s 1. Overlayed are two contours for the 500-hPa high-pass filtered transient eddy geopotential height field as a measure for the storm track activity (solid line is for 50 m, and dash-dotted line is for 70 m). isentropic trajectories and neglect the potential impact of crossisentropic transport) that revealed a distinct transport barrier near the location of the steepest tropopause ( K) in winter [Haynes and Shuckburgh, 2000], a minimum in wintertime STE at K [Morgenstern and Carver, 2001], and a maximum of cross-tropopause exchange above 350 K during summer [Chen, 1995]. This behavior is also confirmed by in situ tracer measurements in the lower stratosphere [Ray et al., 1999], which showed very little mixed-in tropospheric air near 360 K during spring and much larger tropospheric contributions in September on the same isentropic surface Sensitivity to the residence time threshold. [19] To assess the sensitivity of the meridional exchange distributions to the threshold residence time, zonally integrated cross-tropopause mass fluxes are intercompared in Figure 6 for the winter season for t 8 varying between 24 and 96 hours. For the other seasons (not shown) the results are qualitatively similar. As expected from section 3.1, the differences are rather substantial. For example, in the extratropics the zonally integrated STE and TSE mass flux values (Figures 6a and 6b) drop by more than a factor of 3 when changing t 8 from 24 to 96 hours. The sensitivity to t 8 is even more pronounced in the subtropical region (20 30 N) where the mass flux values for t 8 = 96 hours are about 8 times smaller than for t 8 = 24 hours. This clearly reveals the high frequency of tropopause crossings on a relatively short timescale of up to 1 2 days and that the quantitative STE and TSE mass flux estimates strongly depend upon whether these transient events are considered or not. The same is true when considering net (Figure 6c) and two-way (Figure 6d)

8 ACL 13-8 WERNLI AND BOURQUI: EXTRATROPICAL CROSS-TROPOPAUSE EXCHANGE TSE winter TSE spring a) b) TSE summer TSE autumn c) d) Figure 8. Same as Figure 7 but for TSE (and with different contour values). exchange fluxes. For larger residence time threshold values both fluxes are significantly reduced at all s, in particular in the subtropics where the net upward flux becomes 5 times smaller for t 8 = 96 hours compared with the value for t 8 = 24 hours. Note, however, that when integrating meridionally, the net mass flux becomes relatively insensitive to t 8 in the range from 48 to 96 hours (see Figure 2a) Geographical Distribution [20] Next, the detailed geographical distributions of the seasonally integrated STE and TSE mass fluxes (with a threshold residence time of 96 hours) are presented and qualitatively compared with the dynamical storm track activity in the Northern Hemisphere. (The 500-hPa block-filtered transient eddy geopotential height field [Hoskins et al., 1989; Hall et al., 1994] is used as a measure for the dynamic storm track). [21] Maxima of STE (Figure 7) occur within the mid storm track regions over the Pacific and Atlantic oceans, except for summer, where the zonal variability is very weak. Within the storm tracks, there is a lot of variability and no general preference for STE to occur either in the entrance or exit regions. Another area with enhanced STE activity are the Asian mountains during winter and spring. Other preferred continental regions for STE throughout the year are western North America and the area to the west and southwest of the Alps. [22] TSE, on the other hand (Figure 8), has pronounced maxima in the subtropics and in the northern parts of the mid storm tracks. Within the subtropical belt, there is substantial seasonal variability with maximum activity over the Atlantic during winter. Notable regions with enhanced TSE activity in the mids are northeastern Canada during all seasons and the eastern Mediterranean in summer and autumn. [23] It should be pointed out that for the detailed geographical distribution of STE and TSE the interannual variability might be significant and that a priori the details of the identified features for can not be regarded as representative in a climato-

9 WERNLI AND BOURQUI: EXTRATROPICAL CROSS-TROPOPAUSE EXCHANGE ACL STE winter 100 STE summer pressure [hpa] a) b) STE winter 100 STE summer pressure [hpa] c) d) Figure 9. Zonally integrated distribution of (a, b) STE exchange locations and of (c, d) the lowest and highest trajectory points for the winter (left panels) and summer season (right panels) based upon a threshold residence time of 96 hours. Values denote the number of events within a 2 20 hpa large grid box. The solid bold line denotes the position of the seasonal and zonal mean 2-pvu tropopause, and the dashed lines in Figures 9a and 9b indicate the minimum and maximum altitude of the zonally varying tropopause position. In Figures 9c and 9d the distribution of the lowest points is shown by shading, and the distribution of the highest points is shown by solid contours (with contour values 100, 250, 400, and 600). logical sense. Nevertheless, the identified patterns reveal significant zonal variability and, in particular during winter, a qualitative relationship with baroclinic wave activity Vertical Penetration of Exchange Air Parcels [24] The vertical distribution of the exchange locations and the highest and lowest points of the 9-day exchange trajectories are shown for the winter and summer seasons in Figure 9 for STE and in Figure 10 for TSE. (The highest and lowest points correspond with no further constraint to the minimum and maximum pressure values along the 9-day trajectories (which are symmetric in time around the time of exchange).) In the extratropics, STE occurs typically hpa below the height of the climatological tropopause (Figures 9a and 9b). This indicates that STE events are frequently associated with tropopause folds which come down as low as 550 hpa in the region N during winter. In the summer season, only very few exchange events occur below 450 hpa. The panels with the lowest and highest trajectory points (Figures 9c and 9d) reveal that STE air parcels typically descend (during the considered time period of 9 days) from 250 hpa near N to 450 hpa near N during winter (and to about 350 hpa during summer). However, during the cold season, some 10 20% of the exchange parcels penetrate to the lowest troposphere, mainly in the subtropical

10 ACL WERNLI AND BOURQUI: EXTRATROPICAL CROSS-TROPOPAUSE EXCHANGE 100 TSE winter 100 TSE summer pressure [hpa] a) b) TSE winter 100 TSE summer pressure [hpa] c) d) Figure 10. Same as Figure 9 but for TSE. s near 30 N. We will refer to events with their lowest points below 700 hpa as deep exchange events and discuss them in more detail in section 3.5. [25] In contrast to STE, mid TSE occurs typically close to or slightly above the averaged tropopause height in all seasons (Figures 10a and 10b). The subtropical events take place at levels from 100 to 200 hpa. It is notable that comparatively few exchange events (STE and TSE) occur in the climatologically steepest tropopause region near 28 N during winter (Figures 9a and 10a). On the considered timescale of a few days the vertical penetration in the stratosphere is limited to a 100-hPa deep layer just above the tropopause (Figures 10c and 10d). This layer corresponds to the mixing layer in the lowermost stratosphere identified by the analysis of in situ tracer measurements [Lelieveld et al., 1997; Fischer et al., 2000; Hoor et al., 2002; Zahn, 2001]. From our analysis it seems that the vertical location and extent of this layer is similar during all seasons. The vertical distribution of the lowest points of the TSE trajectories shows a similar seasonal cycle as for STE (Figures 10c and 10d). During the winter season a substantial part of mid TSE air parcels have their origin below the 500-hPa level and some even in the lowest troposphere ( deep TSE events ). As for STE, subtropical TSE events are generally associated with less vertical transport. During spring and autumn the vertical distributions (not shown) are intermediate to the ones presented for the winter and summer seasons Deep Exchange Events [26] Finally, we further investigate the vertically deep exchange events, that is, the STE events where the originally stratospheric air penetrates to levels below 700 hpa, and the TSE events where originally tropospheric air from below the 700-hPa level is injected into the stratosphere within 4 days. These events potentially have a particularly important chemical impact as they lead to the rapid interaction of (near) boundary layer and stratospheric air, which are characterized by very different chemical compositions. [27] Figure 11 shows the zonally integrated cross-tropopause mass fluxes associated with deep STE and TSE events for the

11 WERNLI AND BOURQUI: EXTRATROPICAL CROSS-TROPOPAUSE EXCHANGE ACL downward mass flux [10 kg km s ] a) deep STE upward mass flux [10 kg km s ] b) deep TSE Figure 11. Zonally integrated cross-tropopause mass fluxes associated with deep exchange events only (with t 8 = 96 hours) for the four seasons from May 1995 through April 1996: (a) STE and (b) TSE. Values are in 10 6 kg km 1 s 1. The solid line is for winter, the long-dashed line is for spring, the short-dashed line is for summer, and the dashed-dotted line is for autumn. four seasons. As indicated qualitatively by Figures 9 and 10, deep STE is characterized by a marked seasonal cycle (much more prominent than when considering all STE events; see Figure 3a) with a maximum in winter near 45 N and a reduction by a factor of 10 and a northward shift of the maximum during summer (Figure 11a). Integrated over the hemisphere, deep events account for about 15% of the total STE mass flux (with t 8 = 96 hours) during winter. For deep TSE events the seasonal variability is much weaker, and they also occur most frequently during the cold season (Figure 11b). Comparison with Figure 3b reveals that a substantial part of the northern mid TSE events are deep (about 20 25%), but almost none of the subtropical ones are. [28] The geographical distribution of deep exchange events is illustrated in Figure 12 for the winter season. During the considered winter of , deep STE and TSE almost exclusively deep STE deep TSE a) b) Figure 12. Geographical distribution of the exchange mass flux during winter associated with only deep exchange events for (a) STE and (b) TSE. As in Figures 7 and 8, the threshold residence time is 96 hours, the values are in kg km 2 s 1, and the overlayed two contours denote the dynamical storm track activity.

12 ACL WERNLI AND BOURQUI: EXTRATROPICAL CROSS-TROPOPAUSE EXCHANGE deep STE destinations deep TSE origins a) b) Figure 13. Geographical distribution of the (a) destinations of deep STE and (b) origins of deep TSE events during winter. Values correspond to the total number of deep exchange trajectory points below 700 hpa within a 3 3 grid box. As in Figures 7, 8, and 12, the overlayed two contours denote the dynamical storm track activity. occur within (and slightly downstream of) the dynamical storm track regions and for deep TSE also near the Pole. Deep STE has distinct maxima in the preferred areas for cyclogenesis near the Asian and American east coasts (which, according to Elbern et al. [1998], are also preferred regions of tropopause folding) and in the southern parts of the storm track exit regions over California and the Mediterranean Sea. For spring and autumn the qualitatively close correlation with the storm track regions persists (not shown). During summer a few deep STE events occur over the American west coast and Scandinavia, whereas deep TSE is localized in the central North Pacific and near Greenland (not shown). [29] Figure 13 indicates, again for the winter season only, where deep STE air parcels can be observed below 700 hpa during the 4 days after the exchange ( destination of deep STE, Figure 13a) and from which locations below the same level the air is rapidly transported into the stratosphere during the following days ( origins of deep TSE, Figure 13b). (The technical procedure is such that during the 4 days after the exchange it is verified every 6 hours whether the trajectory is below 700 hpa. Therefore a single deep STE trajectory can contribute with several low points ( p > 700 hpa) to Figure 13a. The same procedure is applied to obtain the distribution shown in Figure 13b.) These diagrams provide important information as they point to the regions where, for instance, stratospheric ozone intrusions can be observed near the ground (deep STE), or from where anthropogenic pollutants are likely to be conveyed rapidly into the stratosphere (deep TSE). For the considered winter the destination points of deep STE events (Figure 13a) are preferentially along the southern border of the storm tracks and in continental areas to the south of their exit regions (Baja California, Mediterranean, and northern Africa). On the other hand, the origins of deep TSE events (Figure 13b) reveal two wellconfined prominent hot spots in the southwestern parts of the Atlantic and Pacific storm tracks and almost zero values over the Eurasian continent. 4. Discussion and Further Remarks [30] The following sections contain a detailed discussion of the limitations and value of the utilized Lagrangian approach (sections 4.1 and 4.2), an interpretation of some of the results from the standpoint of dynamical meteorology (section 4.2), and an outlook for further investigations (section 4.3) Caveats of the Methodology [31] There are also limitations associated with the applied methodology and the available atmospheric data sets. Possibly the largest one is related to deep convection which plays an important role for rapid vertical transport in the extratropics (especially during summer) but is not explicitly resolved by the ECMWF model. As a consequence, our deep exchange estimates do not contain the effects of cumulus convection. This might lead to an underestimation of deep exchange, in particular for the summer season. On the other hand, it should be noted that the use of assimilated data guarantees that in regions where vertical temperature soundings are available, at least the influences of cumulus clouds on the atmospheric stability and the height of the tropopause are captured by the approach. [32] A second important caveat is related to the accuracy of the trajectories over a time period of almost 10 days and the calculation of the PV values along these trajectories. Here the most critical parameters are the temporal resolution of the ECMWF analyses (6 hours) (during this time, air parcels near the tropopause can travel over distances of about 1000 km) for the trajectory calculations and the vertical data resolution (11 levels between 100 and 400 hpa) for the accurate determination of the 2-pvu tropopause. The limited temporal data resolution and the necessity of temporal interpolation of the wind fields induce trajectory position errors, which are difficult to quantify. For detailed case studies, mesoscale model simulations can provide wind fields with an increased resolution but only for a limited time period and spatial domain. On the other hand, for a hemispheric climatology of cross-tropopause exchange, ECMWF analysis data constitute a state-of-the-art data set with maximum benefit from the globally available observations. In addition, it is noted that (1) our Lagrangian approach with the exclusion of exchange events with very short residence times effectively eliminates spurious tropopause crossings [see also Bourqui, 2001] and (2) previous comparisons of trajectory ensembles calculated from ECMWF (1 ; 6 hours) and mesoscale model data (0.5 ; 1 hour) yielded satisfactory agreement [Wernli, 1997, Appendix].

13 WERNLI AND BOURQUI: EXTRATROPICAL CROSS-TROPOPAUSE EXCHANGE ACL Residence Time Concept and Deep Exchange [33] Despite these limitations, the applied Lagrangian methodology reveals novel and fundamental characteristics of stratosphere-troposphere exchange in the extratropics, which could not be attained by a purely Eulerian approach. It confirms and extends results from our preliminary investigations [Wernli and Bourqui, 1999] and from the independent Lagrangian analysis of Stohl [2001]. First, the method provides detailed information about the horizontal and vertical locations of exchange events. In the vertical, STE occurs in a 150-hPa deep layer below (and TSE slightly above) the climatological height of the tropopause, as qualitatively pointed out by Juckes [2000] (see discussion of his Figure 5). In the horizontal the Atlantic and Pacific storm tracks are preferred regions for downward cross-tropopause exchange, especially during the cold season. Secondly, it was shown that some exchange events are associated with substantial vertical transport, in particular during winter. From a chemical perspective, vertically deep exchange events (where air is transported from the stratosphere to the potentially polluted atmospheric boundary layer (or vice versa) within a couple of days) are highly important. Third, the trajectory-based analysis showed that the major part of the air parcels which cross the tropopause remain less than 1 2 days in the new sphere before they return to their original one. These events presumably have only little chemical impact because of the limited probability for mixing and chemical reaction within the new environment. The elimination of these transient cross-tropopause exchange events with a residence time threshold criterion has a significant impact upon quantitative mass flux estimates and indicates that quantitative mass flux estimates depend upon the considered timescale. Furthermore, it is not sufficient (from a chemical perspective) to consider only the net cross-tropopause exchange fluxes but also the cancelling (from a mass perspective) two-way exchange, that is, the upward and downward exchange that occurs with equal magnitude within a certain bin. [34] These are important findings which rely on the knowledge of the time history and future of the exchange air parcels and which underpin the value of the adopted Lagrangian approach. The present analysis focused on the climatological aspects of crosstropopause mass exchange. Further complementary investigations are needed to also improve the understanding of the physical and dynamical processes that lead to the exchange (see section 4.3). [35] The presented results demonstrate that STE takes place preferentially below the height of the climatological tropopause, that is, within transient systems with a low tropopause. This qualitatively confirms the importance of tropopause folds and dry intrusions [e.g., Danielsen, 1968; Shapiro, 1980; Browning and Reynolds, 1994; Wernli and Davies, 1997; Stohl and Trickl, 1999], in particular for deep STE. Note also that the minimum of (deep) STE in summer is qualitatively in agreement with the seasonal cycle of tropopause fold activity [Elbern et al., 1998]. In contrast, TSE mainly occurs in regions with a slightly elevated tropopause. Of particular interest are deep TSE events because their source regions coincide with typical boundary layer entrance regions of warm conveyer belts (or moist ascending airstreams) associated with extratropical cyclones [Wernli and Davies, 1997; Stohl, 2001]. There is evidence from in situ and Lidar observations [Arnold et al., 1997; Bethan et al., 1998; Stohl and Trickl, 1999] that these airstreams can transport strongly polluted air to the upper troposphere. They are associated with strong ascent, cloud diabatic effects, and the diabatic production of relatively moist low-pv anomalies just below the elevated tropopause [Wernli, 1997; Pomroy and Thorpe, 2000; Liniger and Davies, 2001]. The present analysis indicates that these areas can be associated with TSE after the ascending phase of the conveyor belt flow, in agreement with the observational analysis of Zahn [2001] based upon isotope measurements and the case study examples of Rood et al. [1997] and Hoor et al. [2002]. The idealized study of Zierl and Wirth [1997] considered the radiative effects on moist anomalies below an elevated tropopause and found significant TSE. Further investigations using real data could be helpful to assess the relative importance of radiative and turbulent processes for these exchange events Future Investigations [36] The discussion of the findings from the present study is concluded with suggestions for future investigations, which could help to assess the representativity of this 1-year climatology and to extend the understanding for the physical and dynamical processes associated with cross-tropopause transport. [37] Open questions are, for instance, the relative importance of turbulence, radiation and condensational heating for the nonconservation of PV along the exchange trajectories (which might be seasonally and geographically dependent), and the predominant dynamical features near the tropopause level that accompany the exchange events (as outlined in section 1). The second question has been approached in an objective way [Sprenger et al., 2001] and preliminary results for a selected month indicate that more than 50% of the exchange events in the extratropics occur in the vicinity of PV streamers and cutoffs. Work is in progress to further analyze this relationship and also the link with other synoptic and subsynoptic scale dynamical features like cyclones, anticyclones and tropopause folds. Considering the occurrence of cross-tropopause transport with different residence times, it is of central importance to assess the spectrum of timescales involved in the small-scale mixing processes [Esler et al., 1999]. [38] The availability of extended reanalysis data sets allows repetition of the adopted methodology for longer time periods and to investigate also the interannual variability of cross-tropopause exchange. The study by Langford [1999] indicates, for instance, a strong sensitivity of cross-tropopause exchange in the eastern Pacific to the El Niño Southern Oscillation phase, and it is conceivable that in the Atlantic region, similar modulations of the frequency and geographical distribution of the exchange occur in harmony with the North Atlantic Oscillation. An extension of the present climatology using the ECMWF 15-year reanalysis data set is underway, and preliminary results are given by Sprenger et al. [2001]. [39] Such a more robust climatology could also be used by chemical transport models to parameterize the cross-tropopause transport [e.g., Lamarque et al., 1999]. From the present analysis it is already clear that for an accurate representation of the exchange mass fluxes it would be essential to take into account the distinct geographical patterns, the characteristic seasonal cycles, the difference in the typical exchange altitude for STE and TSE, and the sensitivity to the residence time. This constitutes a significantly more complex problem than the stipulation of zonally uniform exchange flux values at the height of the climatological tropopause. [40] Finally, it is rewarding to validate the method and its results for selected case studies with other data sets (for instance, with output from mesoscale simulations [cf. Bourqui, 2001]), methodological approaches, and surface and upper-air observations (for instance, in situ aircraft measurements or ozone soundings). Of particular interest is the comparison with a particle dispersion model [Stohl et al., 2000], where, in contrast to a pure trajectory tool, the effects of turbulent mixing and deep convection can be approximatively included via parameterizations. Such a model intercomparison study is underway within the EU project STAC- CATO (influence of stratosphere-troposphere exchange in a changing climate on atmospheric transport and oxidation capacity). 5. Conclusions [41] A fully Lagrangian methodology has been used to derive a 1-year climatology of cross-tropopause mass fluxes in the extra-